Alex Kudriashova Interview
summary
Platform handles 5–6 petabytes of satellite imagery growing ~500 TB/year, so processing targets only requested windows to avoid moving whole scenes. First-day traffic spikes of tens of thousands force on-demand scaling, achieved with AWS Lambda microservices running up to 10,000 functions in parallel. That pipeline builds a Corn Belt crop map in about 30 minutes, supports daily 22 m coverage and planned 2.5 m weekly imaging, and yields fast iterations for new domains.
AI-written summary · about
related talks
20mOpen-source processing pipelines for Cyclone Gabrielle
36mRapid Processing of NASA Satellite Data Stored with Ceph
39mSatellite Imagery to Machine-Readable Intelligence
44mBest practices for analytics and generative AI on AWS(ANT302)
01h 04mBuilding Efficient and Frugal Image Data Processing Pipelines
17mIt's All in the Cloud
51mWhen the Universe Calls
19mEstablishing a Serverless Dynamic Tile Generation Workflow